Machine Translation / Computer-assisted Human Translation
Machine translation has achieved successes; especially in the last decade, statistical machine translation (SMT) systems have replaced some human labors in practical workflows and neural machine translation (NMT) systems even surpass them. However, it still cannot meet the strict requirements for high quality translations in real-world applications. We are working on improving NMT and SMT systems from various aspects. Besides, we are also exploring practical means of using SMT/NMT and other NLP technologies to improve productivity of human translation.
Technologies including Machine Translation (MT)
- Exploiting monolingual corpora for unsupervised and weekly-supervised machine translation
- Extraction of parallel sentences from monolingual corpora [Marie+, ACL 2017] [Marie+, NAACL 2019]
- Extraction of phrasal translations from monolingual corpora [Marie+, TACL 2017] [Marie+, TALLIP 2018]
- Unsupervised machine translation [Marie+, WMT 2019] [Marie+, TALLIP 2020]
- Neural machine translation for low-resource scenarios
- Multilingualism and transfer learning [Imankulova+, MT Summit 2019] [Dabre+, EMNLP 2019] [Dabre+, WMT 2020]
- Unsupervised domain adaptation / style transfer [Marie+, TACL 2020]
- Hyper-parameter tuning [Rubino+, MT Journal 2020]
- Translation quality estimation [Liu+, TASLP 2017] [Rubino+, Eval4NLP 2021] [Rubino+, WMT 2021]
- Reranking of multiple MT systems [Marie+, AMTA 2018]
- General methodologies for neural networks [Dabre+, AAAI 2019] [Dabre+, WNGT 2020] [Dabre+, MT Summit 2021]
- Creating datasets for research
- NICT QE/APE: Spoken, Ja->En/Zh/Ko, Quality estimation and post-editing [Fujita+, WAT 2017]
- JaRuNC: Written, News commentary, Ja/Ru/En, machine translation [Imankulova+, MT Summit 2019]
- Staged PE: Written, En->Ja, Post-editing and error annotation [Fujita, MT Summit 2021]
- Meta analysis of machine translation research [Marie+, ACL 2021]
Computer-assisted Human Translation and Translation Studies
- Designing and developing metalanguages for translation process
[Fujita+, TAUS Asia 2019]
[Yamada+, TT5 2020]
- Quality assessment scheme for learner translators [Fujita+, LAW 2017]
- Taxonomizing translation strategies [Yamamoto+, IATIS 2021]
- Analysis of pre-editing operations effective in improving MT [Miyata+, EAMT 2017] [Miyata+, EACL 2021]
- Translation training environment for project-oriented translation scenarios [Kageura+, 2016] [Kageura+, 2017]